Three Isoforms of a Hepatocyte Nuclear Factor-4 Transcription Factor with Tissue- and Stage-specific Expression in the Adult Mosquito*

We cloned three isoforms of hepatocyte nuclear factor-4 (HNF-4) from the mosquito Aedes aegypti, designated AaHNF-4a, AaHNF-4b, and AaHNF-4c. AaHNF-4a and AaHNF-4b are typical members of the HNF-4 subfamily of nuclear receptors with high amino acid conservation. They differ in N-terminal regions and exhibit distinct developmental profiles in the female mosquito fat body, a metabolic tissue functionally analogous to the vertebrate liver. The AaHNF-4b mRNA is predominant during the previtellogenic and vitellogenic phases, while the AaHNF-4a mRNA is predominant during the termination phase of vitellogenesis, coinciding with the onset of lipogenesis. The third isoform, AaHNF-4c, lacks part of the A/B and the entire C (DNA-binding) domains. The AaHNF-4c transcript found in the fat body during the termination of vitellogenesis may serve as a transcriptional inhibitor. Both AaHNF-4a and AaHNF-4b bind to the cognate DNA recognition site in electrophoretic mobility shift assay. Dimerization of AaHNF-4c with other mosquito HNF-4 isoforms or with mammalian HNF-4 prevents binding to the HNF-4 response element. In transfected human 293T cells, AaHNF-4c significantly reduced the transactivating effect of the human HNF-4α1 on the apolipoprotein CIII promoter. Electrophoretic mobility shift assay confirmed the presence of HNF-4 binding sites upstream of A. aegypti vgand vcp, two yolk protein genes expressed in the female mosquito fat body during vitellogenesis. Therefore, HNF-4, an important regulator of liver-specific genes, plays a critical role in the insect fat body.

The insect fat body is a functional analogue of the vertebrate liver. Serving as a "protein factory" (1), it is the insect's most powerful secretory organ, responsible for production of virtually all hemolymph proteins (2). The aptness of this biological analogy is further substantiated by the many parallels in regulatory mechanisms responsible for fat body-and liver-specific gene expression, which is governed by a combination of transcription factors synergistically interacting with regulatory elements in target genes. Indeed, the regulatory sequences of some fat body-specific genes and their liver-specific counterparts are highly conserved (3)(4)(5), and several fat body-specific transcription factors are similar to those found in liver-specific regulatory pathways (6 -9).
Hepatocyte nuclear factor 4 (HNF-4), 1 a member of the nuclear receptor superfamily, is a transcriptional activator for a wide variety of liver genes with diverse functions (10 -12). Apart from directly regulating target genes, HNF-4 may trigger a cascade of transcription factors involved in appropriate expression of genes essential for liver function (10). For example, HNF-4 activates transcription of HNF-1, another important hepatocyte transcription factor (13,14).
We are interested in understanding the molecular mechanisms of insect vitellogenesis. In the anautogenous mosquito Aedes aegypti, this process is initiated by a blood meal and involves the coordinated activity of two vitellogenic tissues: the fat body, which produces and secretes several yolk protein precursors, and the ovary, which specifically accumulates them (15). Vitellogenic events are regulated, at least in part, by the insect steroid hormone, 20-hydroxyecdysone (20E). The expression of vitellogenic genes is mediated through a 20E-triggered regulatory cascade of transcription factors (16), similar to that in Drosophila development (17).
A high degree of evolutionary conservation among HNF-4 sequences suggests that the mosquito homologue may be an important transcription factor in the fat body, as it is in the vertebrate liver. Indeed, vertebrate HNF-4 cloned from rat, mouse, Xenopus, and human (10, 18 -22), as well as their insect counterparts from Drosophila (23) and Bombyx mori (24), show a very high amino acid sequence similarity in the DNA-binding, hinge, and dimerization/ligand-binding domains. The tissue distribution of HNF-4 has been conserved throughout evolution as well. The somatic tissues, where HNF-4 is expressed, are mainly restricted to the vertebrate liver, intestine, and kidney (18,19,25) or to their functional analogues in the Drosophila embryos: the fat body, midgut, and Malpighian tubules, respectively (23). In the silk moth B. mori, HNF-4 expression was detected in most tissues of the larva and pharate adult, with highest expression in the gut, fat body, and gonads (24).
In this paper, we describe the cloning and characterization of three isoforms of mosquito HNF-4: AaHNF-4a, AaHNF-4b, and AaHNF-4c. Three distinct transcripts of 2.8 kb (AaHNF-4a), 2.1 kb (AaHNF-4b), and 1.8 kb (AaHNF-4c) were detected in adult mosquitoes, and their distribution is tissue-and stagespecific. The AaHNF-4a and AaHNF-4b isoforms are typical members of the HNF-4 subfamily, and their functionality has been confirmed by their ability to bind as homodimers to the cognate DNA recognition site. In the mosquito fat body, AaHNF-4b is predominant during the previtellogenic period and the synthetic phase of vitellogenesis, while AaHNF-4a is predominant during the termination phase. In the ovaries, only the AaHNF-4b transcript was detected throughout the entire vitellogenic period. A novel finding of this work is the cloning of a unique mosquito isoform AaHNF-4c; it is distinct from AaHNF-4b in that part of the A/B and the entire C (DNAbinding) domains are absent. The AaHNF-4c transcript appears most abundantly during the termination phase of vitellogenesis. The AaHNF-4c isoform prevented AaHNF-4a, AaHNF-4b, and rat HNF-4 from binding to DNA in EMSA, suggesting that this HNF-4 isoform may serve as a transcriptional repressor. Furthermore, in transfected human embryonic kidney 293T cells, AaHNF-4c significantly reduced positive transcriptional activity of the human HNF-4␣1 on the apolipoprotein CIII promoter (apo-CIIIP). In addition, we also have identified HNF-4 binding sites in the upstream region of two yolk protein precursor genes: vitellogenin (vg) (26), and vitellogenic carboxypeptidase (vcp) (27), which are the major genes specifically expressed in the female fat body during vitellogenesis (15).

MATERIALS AND METHODS
Animals-Mosquitoes, A. aegypti, were reared according to Hays and Raikhel (28). Larvae were fed on a standard diet as described before (29). Vitellogenesis was initiated by allowing females 3-5 days after eclosion to feed on an anesthetized white rat.
Materials-The RNA ladder was purchased from Life Technologies, Inc.; Sequenase was from U. S. Biochemical Corp.; and restriction enzymes were from Boehringer Mannheim. Perkin-Elmer was the source of reagents for the polymerase chain reaction (PCR), and in vitro transcription and translation assays were from Promega. MSI CO supplied nitrocellulose-blotting membranes. Radionucleotides for labeling of probes and DNA sequencing were from NEN Life Science Products. All other reagents were of analytical grade from Sigma or Baker.
Cloning and Sequencing of cDNA-A cDNA fragment of AaHNF-4 was first obtained by PCR, for which degenerate primers were designed based on the sequences of rat, mouse, and Drosophila HNF-4 (18,19,23). Amplification was achieved in a Perkin Elmer thermal cycler using as the template cDNA reverse-transcribed from 20 g of total RNA prepared from the fat bodies of vitellogenic female mosquitoes. The PCR-generated fragment was used as a probe to screen a ZAPII cDNA library, which was prepared from the fat bodies of vitellogenic female mosquitoes 6 -48 h post-blood meal (PBM) as previously reported (30). Several positive cDNA clones were subsequently isolated and sequenced using standard protocols (31). Analyses of nucleotide and deduced amino acid sequences were performed using the software of the University of Wisconsin Genetics Computer Group.
Northern Blot Hybridization-Total RNA was isolated from mosquitoes of different stages and tissues using the guanidine isothiocyanate method as described previously (32). Polyadenylated mRNA was isolated using Biomag oligo(dT) 20 magnetic beads and the manufacturer's protocols (PerSeptive Diagnostics, Inc.). For Northern blot analysis, total or poly(A) ϩ RNA was separated by electrophoresis in 1.2% agarose/formaldehyde gels in MOPS buffer, blotted to a nitrocellulose membrane, and hybridized to 32 P-labeled DNA probes under high stringency conditions (31). Autoradiography was conducted at Ϫ70°C using intensifying screens.
In Vitro Transcription and Translation-The mosquito AaHNF-4a and -4b cDNAs were subcloned into the transcription vector pGEM.3Z (Promega) under the control of the SP6 RNA polymerase promoter. The mosquito AaHNF-4c isoform and the rat HNF-4 clone (a kind gift of Dr. F. Sladek) (18) were subcloned into pBluescript, under the control of the T3 RNA polymerase promoter. The DmGATAb clone was a gift form Dr. T. Abel and was under control of the T7 RNA polymerase promoter (33).
The TNT System (Promega) was used for in vitro TNT of cDNA clones in rabbit reticulocyte lysate, utilizing the corresponding RNA polymerases. 1 g of DNA was used in a total reaction volume of 50 l. The TNT reactions were conducted at 30°C for 2 h and then stored at Ϫ70°C until needed for electrophoretic mobility shift assays.
Electrophoretic Mobility Shift Assay (EMSA)-Binding reactions were carried out in a total volume of 20 l containing 2-3 l of the appropriate TNT sample, 10 mM Tris-HCl (pH 8.0), 50 mM NaCl, 1 mM MgCl 2 , 0.5 mM dithiothreitol, 0.5 mM EDTA, 2 g poly(dI-dC), 4% glycerol, and 5 pmol (100-fold molar excess) of competitor DNA when appropriate. After incubating for 15 min at room temperature, 0.05 pmol of 32 P-labeled DNA probe was added, and the solution was incubated an additional 15 min. The samples were loaded on a 5% nondenaturing polyacrylamide gel (prerun for 1 h) in 0.5ϫ TBE and run at 10 V/cm. The gel was then dried and autoradiographed with an intensifying screen at Ϫ70°C.
To initiate heterodimer formation between AaHNF-4c and other HNF-4 isoforms, the TNT samples were first mixed together and heated at 50°C for 1 min and then cooled at room temperature for 15 min (34,35). The volume was brought up to 20 l in shift buffer, and the reaction was carried out as described above. Control incubations with each HNF-4 isoform without the addition of AaHNF-4c showed no or negligible reduction in specific binding due to heating.
The APF-1 probe was labeled by back-filling with Klenow fragment using [␣-32 P]dCTP, and the Vg and VCP HNF4 probes were labeled by back-filling with [␣-32 P]dATP. The BZIP-1 probe used as a nonspecific competitor and the Box-A probe used as a positive control for Dm-GATAb were described previously (37).
Transient Transfection-The AaHNF-4c cDNA was cloned into the EcoRI site of the expression vector pcDNA3.1/Zeo(ϩ) (Invitrogen). Construction of the human HNF-4␣1 expression vector and the luciferase reporter construct pL854, containing the apo-CIIIP, have been previously described (38). Human embryonic kidney 293T cells were seeded at a density of 1.2 ϫ 10 5 in 12-well tissue culture plates and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Each transfection contained 150 ng of pL854, 50 ng of pCMV␤-gal (CLONTECH), selected amounts of human HNF-4␣1 and/or AaHNF-4c, and empty pcDNA 3.1 vector for a total of 5 g of DNA. Cells were transfected in serum-free Dulbecco's modified Eagle's medium using LipofectAMINE reagent (Life Technologies) at a DNA: LipofectAMINE ratio of 1:4. At 4 h after transfection, the cells were incubated in Dulbecco's modified Eagle's medium plus 10% fetal bovine serum for an additional 36 h. The human liver HepG2 cells were cultured as described previously (36). Luciferase and ␤-galactosidase activity were determined by the Dual-Light reporter gene assay system (Tropix) using a Wallac-Berthold LB96P-2 luminometer.

Isolation and Characterization of cDNA Clones Encoding
Mosquito AaHNF-4 Isoforms-Putative clones encoding the mosquito HNF-4 transcription factor (AaHNF-4) were obtained by a combination of the PCR and cDNA library screening. The following strategy was used to design degenerate primers. The sense primer conformed to the P-box sequence of the DNAbinding domain, which is conserved among the members of the nuclear receptor superfamily; the antisense primer was based on the conserved sequence LDDQVA from the dimerization domain of rat, mouse, and Drosophila HNF-4 homologues (18,20,23). A 440-base pair fragment was amplified by PCR and sequenced as described under "Materials and Methods." After confirming that its sequence was similar to that of rodent and Drosophila HNF-4 isoforms, it was used as a probe to screen a mosquito fat body cDNA library.
Restriction mapping and sequence analyses of positive cDNA clones revealed three distinct mosquito cDNAs present in the library, which we designate AaHNF-4a, AaHNF-4b, and AaHNF-4c ( Fig. 1). The AaHNF-4b (1.86 kb) and AaHNF-4c (1.53 kb) cDNAs share the same sequences at the 5Ј-terminus (except that the 5Ј-end of AaHNF-4c is 23 base pairs longer), whereas the AaHNF-4a (2.43-kb) cDNA has a different 5Јterminal sequence. All three AaHNF-4 cDNAs have the same 3Ј-end sequences and lack a poly(A) tail, indicating that none of the cDNAs represent a full-length transcript. However, AaHNF-4b has a canonical polyadenylation signal (AATAAA) followed by an additional 14 base pairs, while the AaHNF-4a and AaHNF-4c cDNA sequences end just before the position of the AATAAA site in AaHNF-4b. In all three cDNAs, the putative start codons (ATG) are preceded by several in-frame stop codons, indicating that the open reading frame is full-length in each clone.
The conceptual translation of the AaHNF-4a, AaHNF-4b, and AaHNF-4c cDNAs shows that they encode three different proteins, or isoforms, of 565, 538, and 427 amino acids, respectively. The mosquito AaHNF-4a and AaHNF-4b isoforms exhibit a structural domain organization similar to that of the rodent and insect HNF-4 homologues (Fig. 2). The C (DNAbinding), D (hinge), and E (dimerization/ligand-binding) domains are highly conserved; for example, AaHNF-4a and AaHNF-4b share 89, 65, and 79% identity with the respective domains of Drosophila HNF-4 and 91, 58, and 72% identity with B. mori (Fig. 2). An N-terminal A/B and a C-terminal proline-rich F-domain, which possibly have a transactivator function (18), are poorly conserved in the mosquito AaHNF-4, a characteristic feature of all members of the HNF-4 subfamily and, more generally, the nuclear receptor superfamily (40).
A sequence comparison of AaHNF-4a and AaHNF-4b isoforms show that they encode nearly identical polypeptides, differing only in the amino acids at their N termini (Fig. 3), strikingly similar to the B. mori isoforms BmHNF-4a and Bm-HNF-4b. In contrast, the first six amino acids of the shortest isoform, AaHNF-4c, are identical to those of AaHNF-4b, but the rest of the A/B domain and the entire C (DNA-binding) domain are lacking (Fig. 3). The structure of the mosquito AaHNF-4c isoform is unique among insect and vertebrate members of the HNF-4 subfamily that have been reported to date.
Analysis of AaHNF-4 Transcripts-To determine if the three AaHNF-4 cDNAs we have cloned represent respective AaHNF-4 mRNA species in the mosquito, Northern hybridizations were performed using particular portions of the cDNAs as probes (Fig. 4); a probe common to all three cDNAs (Fig. 1, common) detected three transcripts of 2.8, 2.1, and 1.8 kb; a probe specific to the 5Ј-untranslated region of the AaHNF-4a cDNA (5Ј-HNF-4a, Fig. 1) hybridized only with the 2.8-kb transcript; and the last probe, containing sequences from the A/B and C domains common to the AaHNF-4a and AaHNF-4b cDNAs but not to the AaHNF-4c (Fig. 1, ⌬C), recognized the 2.8-and 2.1-kb mRNAs but not the 1.8-kb mRNA. Thus, we differentiated the 2.8-, 2.1-, and 1.8-kb mRNA species as AaHNF-4a, AaHNF-4b, and AaHNF-4c transcripts, respectively (Fig. 4). This experiment verifies that one of the transcripts indeed lacks the region encoding for a portion of the A/B domain and the entire DNA-binding domain, and therefore, AaHNF-4c does not appear to be an artifact of the cDNA library.
To further confirm the presence the AaHNF-4c transcript in the mRNA pool from the mosquito fat body, we performed reverse transcription-PCR using nondegenerate primers that flanked the missing region of AaHNF-4c. cDNA was reverse transcribed from total RNA extracted from the fat bodies of female mosquitoes at 36 h PBM. The sense and antisense primers were designed based on the 5Ј-untranslated region and hinge domain, respectively, common to both AaHNF-4b and AaHNF-4c cDNAs. Upon primary amplification by PCR, only a band of the expected size for AaHNF-4b was detected. However, when a secondary PCR was performed, a band of the expected size for AaHNF-4c was amplified. Southern blot analyses demonstrated that both bands hybridized with a probe specific to the hinge region of HNF-4 (data not shown).
The Distribution of AaHNF-4 Expression in Tissues of the Adult Mosquito-To determine the distribution of AaHNF-4 expression in tissues of adult mosquitoes, polyadenylated mRNA was isolated from the whole body of males and the fat body, midgut, Malpighian tubules, ovary, and thorax of females, taken at the middle stage of the vitellogenic cycle (24 h PBM) and subjected to Northern blot analysis. A fragment of the cDNA common to all three AaHNF-4 cDNAs ( Fig. 1) was used as a hybridization probe. The mosquito fat body, Malpighian tubules, and midgut exhibited significant levels of AaHNF-4 transcription. Two mosquito HNF-4 isoforms (AaHNF-4a and AaHNF-4b) were detected in these tissues (data not shown).
In contrast to the somatic tissues of mosquito females, only the 2.1-kb (AaHNF-4b) transcript was found in the ovaries (Fig. 5). The highest level of ovarian AaHNF-4b mRNA was detected soon after initiation of vitellogenesis by a blood meal (6 h PBM); it dropped to an undetectable level by 36 h PBM, the time of termination of vitellogenic events in the female mosquito (Fig. 5).
Stage-specific Expression of AaHNF-4 Transcripts in the Fat The shaded box in AaHNF-4a shows the 5Ј-region that differs from those of the AaHNF-4b and AaHNF-4c cDNAs. The GAP box indicates that the AaHNF-4c cDNA lacks the corresponding nucleotide sequences relative to the AaHNF-4a and AaHNF-4b cDNAs.
The following cDNA fragments were used in Northern blot analyses as hybridization probes: (i) the 0.3-kb EcoRI-XbaI fragment (5ЈHNF-4a) is specific to the 5Јuntranslated region of the AaHNF-4a cDNA; (ii) the 0.7-kb HindIII fragment (common) contains sequences from the E (dimerization/ligand-binding) and F domains and is common to all three cDNAs; (iii) the 0.3-kb ⌬C-probe, generated by AluI digestion of a 5Ј-deleted AaHNF-4b cDNA, is from the A/B and C (DNA-binding) domains, common to AaHNF-4a and AaHNF-4b cDNAs but not to AaHNF-4c.
Body-The high level of conservation between the mosquito HNF-4 and its vertebrate counterparts (Figs. 2 and 3) as well as the presence of three AaHNF-4 transcripts in the adult fat body (Figs. 4 and 5) suggests that it plays an important role in this metabolic tissue, similarly to that of the vertebrate HNF-4 in the liver. In an attempt to further understand the possible functions of AaHNF-4 in the vitellogenic fat body of adult mosquitoes, we characterized its expression in more detail.
Total RNA from the fat body of female mosquitoes at different stages of vitellogenesis was analyzed by Northern blot hybridization using the cDNA probe common to all three AaHNF-4 isoforms. The expression pattern of the three AaHNF-4 transcripts changed differentially over the course of the vitellogenic cycle (Fig. 6). In the previtellogenic stage, all three transcripts were clearly present during the first day posteclosion, but in subsequent days only the 2.1-kb transcript (corresponding to the AaHNF-4b isoform) was present. Furthermore, its levels increased by day 5 of the previtellogenic period. During the first 18 h of vitellogenic period, only the AaHNF-4b transcript was clearly detectable. At 24 h PBM, when yolk protein gene transcription is nearing its maximum (15), the levels of the AaHNF-4a transcript began to rise again, relative to AaHNF-4b, and the AaHNF-4c transcript appeared. By 36 h PBM, the levels of HNF-4 mRNA significantly increased; AaHNF-4a became the predominant form and AaHNF-4c more pronounced. This pattern of AaHNF-4 expression was maintained until 48 h PBM, when yolk protein synthesis terminated and the fat body returned to its previtellogenic state (15).
DNA Binding by AaHNF-4 Proteins-We used EMSA to determine if the mosquito HNF-4 homologue is a functional DNAbinding protein. The three AaHNF-4 cDNAs were subcloned into transcription vectors and then subjected to TNT reactions in a rabbit reticulocyte lysate system. The results of the TNT reaction were verified by utilizing [ 35 S]methionine; SDS-polyacrylamide gel electrophoresis analyses showed that the in vitro synthesized proteins closely corresponded to their expected molecular sizes (data not shown).
The TNT expressed proteins were examined for their ability to form specific binding complexes with APF-1, a sequence previously shown to be a recognition site for both rat and insect HNF-4 factors (18,23,24). A rat HNF-4 clone was used as a positive control. The AaHNF-4a and AaHNF-4b proteins formed DNA-protein complexes of similar mobility, which migrated slower than the complex formed by the rat HNF-4 (Fig.  7). The binding of these proteins was sequence-specific and could be competed away with an excess of cold probe (APF-1) but not with a nonspecific competitor. The protein synthesized from the AaHNF-4c cDNA, which is missing the DNA-binding domain, failed to exhibit any binding activity (data not shown).
An APF-1 binding complex of intermediate mobility was observed in the retardation gel when the AaHNF-4b isoform (538 amino acids long) and the rat HNF-4 (427 amino acids long) were co-expressed in vitro, suggesting that they formed a mosquito-rat HNF-4 heterodimer (not shown). The retarded bands were completely eliminated by an excess of cold APF-1 sites, indicating that binding was sequence-specific. These data are in agreement with the demonstration by Zhong et al. (23) that the Drosophila and rat HNF-4 proteins can heterodimerize in vitro.
The AaHNF-4c Isoform Is a Transcriptional Repressor-The AaHNF-4c isoform is unique because it lacks the DNA-binding domain. The appearance of its transcript in the fat body during the termination phase of vitellogenesis suggests that it may serve as a repressor by forming inactive heterodimers with other AaHNF-4 isoforms or possibly with other dimeric transcription factors. Utilizing EMSAs, we examined the ability of AaHNF-4c to prevent the other AaHNF-4 isoforms from bind- ing to an HNF-4 response element. To allow heterodimer formation between AaHNF-4c and the other HNF-4 isoforms, the appropriate TNT samples were first mixed together and heated and then allowed to cool prior to adding the radiolabeled DNA probe as described under "Materials and Methods." Incubations of each HNF-4 isoform in the absence of AaHNF-4c showed that heat treatment caused little or no reduction in specific binding (Fig. 8A, lanes 1, 3, and 5). The addition of AaHNF-4c to AaHNF-4a, AaHNF-4b, and rat HNF-4 prevented the formation of binding complexes with the probe (Fig. 8A,  lanes 2, 4, and 6). This inhibition by AaHNF-4c was dose-dependent and specific (Fig. 8B, lanes 2-5). In the control experiment, the addition of increasing amounts of an unrelated transcription factor (Drosophila GATAb) did not interfere with the ability of AaHNF-4b to form the retardation complex (Fig.  8B, lanes 6 -8). Thus, EMSA analyses suggest that the AaHNF-4c isoform may serve as a transcriptional repressor in the mosquito fat body by forming inactive heterodimers with FIG. 4. Northern hybridization analyses of AaHNF-4 transcripts. Polyadenylated mRNA was isolated from the fat body of female mosquitoes at 26 h PBM. Fragments of either AaHNF-4a or AaHNF-4b cDNAs (Fig. 1) were used for Northern blot analyses as hybridization probes: (i) common, a probe common to all three mosquito cDNAs; (ii) ⌬C, a probe common to AaHNF-4a and AaHNF-4b cDNAs, containing sequences from the A/B and C domains; (iii) 5ЈHNF-4a, a probe specific to the 5Ј-untranslated region of the AaHNF-4a cDNA. other AaHNF-4 isoforms.
Transcriptional repression by the AaHNF-4c isoform was further studied using the transient transfection assay in cell culture. We utilized two mammalian cell lines: human hepatic HepG2 cell line, containing endogenous HNF-4 (38) and human embryonic kidney 293T cell line, considered to be devoid of endogenous HNF-4 (41). The reporter construct consisted of the luciferase gene fused to the human apo-CIIIP, containing an HNF-4 response element (38). In pilot experiments, full-length AaHNF-4a and AaHNF-4b failed to activate or repress the human apo-CIIIP in both mammalian cell lines (data not shown) This is most likely due to incompatibility of mosquito HNF-4 with mammalian co-factors. Therefore, we tested the effect of the mosquito HNF-4c isoform on the transcriptional activity of the human HNF-4␣1. In the 293T cells, the human HNF-4␣1 caused a dose-dependent activation of the apo-CIIIP, while transfected mosquito HNF-4c elicited no change in activity (Fig. 9A). When the human HNF-4␣1 and mosquito HNF-4c were co-transfected, apo-CIIIP activity decreased in a dose-dependent manner (Fig. 9B). In HepG2 cells, in which the apo-CIIIP activity is maintained by endogenous HNF-4 (38), transfection of increasing amounts of AaHNF-4c resulted in a dosedependent reduction of this transcriptional activity (data not shown). Thus, AaHNF-4c inhibits transcriptional activation by human HNF-4␣1 most likely by forming heterodimers incapable of DNA binding, thereby causing a dominant negative effect (39).

Identification of HNF-4 Binding Sites in Regulatory Regions of Fat Body-specific, Yolk Protein Precursor
Genes-To further substantiate the role of AaHNF-4 in vitellogenic events in the mosquito fat body, we examined the regulatory regions of two genes encoding yolk protein precursors for HNF-4 binding sites. In A. aegypti, vg (26) and vcp (27) are the two major genes that are expressed in the fat body of blood-fed females (15). Examination of the 5Ј-regions of both A. aegypti vg and vcp genes revealed the presence of putative HNF-4 binding sites, Ϫ351 to Ϫ327 base pairs (Vg HNF4) and Ϫ442 to Ϫ417 base pairs (VCP HNF4) upstream of their transcription start sites, respectively (Fig. 10A). We tested both putative binding sites (via EMSAs) for the ability to compete with the radiolabeled APF-1 sequence for binding by AaHNF-4b. Both VCP HNF4 and Vg HNF4 exhibited significant specific competition, although binding was weaker than that of APF-1; interestingly, VCP HNF4 was a stronger competitor than Vg HNF4 (Fig.  10B).
In a second set of experiments, both VCP HNF4 and Vg HNF4 were tested for direct binding to AaHNF-4b. Both binding sites formed retardation complexes with AaHNF-4b, which were specifically competed by unlabeled APF-1, VCP HNF4, and Vg HNF-4 ( Fig. 10C and data not shown). These tests also showed that the binding affinities of these sites were APF-1 Ͼ VCP HNF4 Ͼ Vg HNF4 (Fig. 10C, lanes 3-5). The addition of a nonspecific competitor had no effect on the retardation complexes (Fig. 10C, lane 6). DISCUSSION In this work, we report the cloning and characterization of the mosquito homologue to HNF-4, a transcription factor that is a member of the nuclear receptor superfamily (10). Significantly, we have identified three mosquito isoforms, designated AaHNF-4a, AaHNF-4b, and AaHNF-4c, and have examined the tissue distribution of expression, focusing on their developmental kinetics in the two vitellogenic mosquito tissues, the fat  7. Electrophoretic mobility shift assay analysis of mosquito HNF-4a and HNF-4b. HNF-4 proteins were synthesized in vitro using a TNT kit. The HNF-4 binding site in the apolipoprotein C III gene (18) (APF-1; TGGGCAAAGGTCA) was radiolabeled and used to test for binding. The isoforms used are noted above each lane. Competitor DNA was added at a 100-fold molar excess to labeled probe. Unlabeled APF-1 was used as a specific competitor; BZIP-1 (ATTTTGCAAT) (37), which binds the CCAAT/enhancer-binding protein, was used as a nonspecific competitor. body and ovary. Moreover, we have found HNF-4 binding sites in two vitellogenic genes that are expressed exclusively in the mosquito female fat body, an insect tissue functionally equivalent to the vertebrate liver.
Comparisons of the deduced amino acid sequences revealed that the AaHNF-4a and AaHNF-4b isoforms are typical members of the HNF-4 subfamily. They differ between themselves only in the N-terminal end of the variable A/B domain. The other domains are identical and show a remarkable level of conservation: the C (DNA-binding), D (hinge), and E (dimerization/ligand-binding) domains share over 90, 77, and 85% FIG. 8. AaHNF-4c inhibits DNA binding by other HNF-4 isoforms. EMSA was used to examine the ability of AaHNF-4c to abolish the DNA binding ability of other HNF-4 isoforms via the formation of heterodimers. All proteins were synthesized in separate TNT reactions. The proteins present in each reaction are noted above each lane. The samples were heated at 50°C for 1 min to allow dimers to break and reform. A, EMSA showing that the addition of AaHNF-4c abolishes the DNA binding ability of AaHNF-4a, AaHNF-4b, and rat HNF-4. B, titration of the effect of AaHNF-4c. A constant amount of HNF-4b was mixed with increasing amounts of either AaHNF-4c or DmGATAb (the total volume was kept constant with unprogrammed lysate). DmGATAb (originally named box A binding factor (33)) was used as a nonspecific competitor for dimer formation. To independently demonstrate the DNA binding ability of DmGATAb, the box A DNA site (33) was used as a probe in EMSA with DmGATAb (lane 9).

FIG. 9. AaHNF-4c inhibits transcriptional activation by human HNF-4␣1.
A, human embryonic kidney 293T cells were transfected with selected amounts of either the human (hHNF-4␣1) or mosquito (AaHNF-4c) isoforms. The total amount of DNA transfected was kept constant at 5 g using empty vector (see "Materials and Methods"). Control is transfection with reporter constructs and empty vector only. Luciferase activity was normalized for transfection efficiency with ␤-galactosidase, and the control sample was set at a value of 1. Each bar represents the mean of three replicates plus the S.E. The human HNF-4␣1 values are significantly different from the control (p Յ 0.05) while AaHNF-4c did not differ significantly. B, co-transfection of human HNF-4␣1 and AaHNF-4c. Cells were co-transfected with human HNF-4␣1 (1 ng) and AaHNF-4c (0, 100, or 250 ng). The total amount of DNA transfected was kept constant at 5 g using empty vector. Luciferase activity was normalized for transfection efficiency with ␤-galactosidase and reported as a percentage of the activity due to human HNF-4␣1 alone. Each bar represents the mean of three replicates plus the S.E. Co-transfection of human HNF-4␣1 with either 100 or 250 ng of AaHNF-4c resulted in a significant decrease in luciferase activity (p Յ 0.05). amino acid similarity with the corresponding domains in rodent and insect HNF-4 isoform (Fig. 2). Furthermore, the AaHNF-4a and AaHNF-4b isoforms are functional DNA-binding proteins, as confirmed by EMSA experiments in which they bind to the cognate DNA recognition site. The third mosquito isoform, AaHNF-4c, is unique among vertebrate and insect members of the HNF-4 subfamily. The greater part of its A/B domain and the entire DNA-binding domain are absent, and consequently, AaHNF-4c cannot bind DNA. In the fat body of adult mosquito females, Northern hybridization revealed three mRNAs of 2.8 kb (AaHNF-4a), 2.1 kb (AaHNF-4b), and 1.8 kb (AaHNF-4c). Utilization of a domain-specific probe in Northern analysis and of reverse transcription-PCR further confirmed the existence of a transcript that indeed lacks the sequence that encodes for a section of the A/B and the entire DNAbinding domains.
Many nuclear receptors have multiple isoforms that originate by different mechanisms (42)(43)(44)(45)(46)(47)(48)(49). While only one HNF-4 isoform has been cloned from Drosophila (23), two isoforms have been detected in both B. mori and mouse (19,24) and four in humans (21,22). In addition, two distinct HNF-4 genes have been detected in both humans and Xenopus (22). It seems likely that a single AaHNF-4 gene encodes the three mosquito isoforms. Comparison of the three mosquito sequences suggests that they are created by usage of unique 5Ј-exons (one encoding the 33 initial amino acids in AaHNF-4a and another the 6 initial amino acids in the AaHNF-4b and AaHNF-4c isoforms) that are alternatively spliced to a common set of 3Ј-exons (Fig.  3). An additional confirmation of the involvement of the mechanism of alternative splicing comes from structural analysis of the mouse HNF-4 gene. It consists of at least 10 exons, with seven of these (exons 2-8) encoding the conserved sequences of the DNA-binding, hinge and dimerization domains (19). Significantly, alignment of the splice junctions of the mouse HNF-4 gene exons with the mosquito sequence shows that the splice junction between mouse exons 3 and 4 falls exactly at the C-terminal end of the "gap" in the AaHNF-4c isoform (shown by an arrow marked 3/4 in Fig. 3). Amplification of genomic DNA by PCR using primers designed on sequences flanking this putative exon junction identified an intron of approximately 3 kb in the same position in the mosquito sequence (data not shown). However, the same approach was unsuccessful when applied to the putative splice junction at the N terminus of the A/B domain (Fig. 3, vertical arrow), probably due to the presence of a very large intron, which is a common feature of genes encoding for nuclear receptors (19,25,42,44,50).
In the adult rat and mouse, the expression of the HNF-4 transcription factor is not restricted to the liver but is fairly high in the kidney and intestine as well (10). Likewise, in the Drosophila embryo, HNF-4 mRNA has been detected in the developing fat body (liver analogue), Malpighian tubules (kidney analogue) and midgut (intestine analogue) during organogenesis (23). HNF-4 transcripts were detected in several tissues of B. mori with the highest levels in the fat body, gut, and gonads (24). In the adult mosquito, in addition to the fat body, both AaHNF-4a and AaHNF-4b are present in the somatic tissues of the midgut and Malpighian tubules. In the ovary, HNF4 (lanes 2-4) or Vg HNF4 (lanes 5-7). Lane 8, competition with a nonspecific probe, BZIP-1. The amount of competitor DNA used (in molar excess to the labeled probe) is noted above each lane. C, EMSA for direct binding of VCP HNF4 site. The VCP HNF4 sequence shown in A was labeled and tested for binding to AaHNF-4b. Lane 1, lysate only; lane 2, binding to in vitro synthesized AaHNF-4b; lanes 3-5, competition by the specific binding sites APF-1, VCP HNF4, and Vg HNF4 (100-fold molar excess); lane 6, competition by a nonspecific probe, BZIP-1 (100-fold molar excess). The HNF-4 binding site consensus sequence is from Sladek (10). Nucleotides in boldface type represent sequences similar to the HNF-4 half-sites. The arrows denote the orientation of the half-sites. Numbers below the sequences indicate the distance from the transcription start sites. K is G or T; R is A or G. B, competition EMSA with HNF-4 binding sites from A. aegypti vg and vcp. HNF-4b protein, synthesized in vitro using a TNT kit, was bound to 32 P-labeled APF-1 probe (lane 1). Competition was performed by adding increasing amounts of unlabeled VCP however, AaHNF-4b is the only transcript detected during the entire vitellogenic cycle. In contrast, two HNF-4 isoforms have been found in the ovary and developing follicles of B. mori (24).
In the fat body of female mosquitoes, the expression pattern of the AaHNF-4 transcripts varies significantly over the course of the vitellogenic cycle (Fig. 6). Importantly, these changes in AaHNF-4 expression correspond to essential shifts in the functional stage of the female mosquito fat body (15). Developmental profiles of the AaHNF-4a and AaHNF-4b transcripts in the adult fat body suggest that these AaHNF-4 isoforms execute distinct functions required at different stages of the vitellogenic cycle. It is noteworthy that in nuclear receptors, the A/B domain has been implicated in both transactivation and recognition of regulatory sequences of various target genes (51).
The AaHNF-4a transcript is present in the female fat body during the first day after eclosion, but its level dramatically decreases in the following days of previtellogenic development, a time when the fat body is preparing for vitellogenesis under the control of juvenile hormone. The AaHNF-4a transcript is practically absent during the first 24 h PBM, during which vitellogenesis is initiated and the expression of yolk protein genes reaches maximal levels under the control of the rising titer of the insect steroid hormone 20E. The AaHNF-4a transcript reappears at about 24 h PBM, and its level gradually rises during the next 12 h, the time when the titer of 20E drops sharply, and the expression of the yolk protein genes is halted. However, a significant elevation in the AaHNF-4a transcript level is observed after 36 h PBM, a time when the synthesis of yolk proteins is terminated and the fat body undergoes remodeling from a protein synthesizing tissue to a storage depot for lipid reserves. Van Heusden et al. (52) reported a 3-fold increase of lipophorin, an insect lipid career protein synthesized by the fat body, in the mosquito hemolymph 40 h after blood feeding. Thus, it is possible that AaHNF-4a may play a role in regulating genes involved in lipogenesis in the mosquito fat body. This suggestion is further supported by a recent report that fatty acyl-CoA thioesters are ligands of HNF-4, which was previously considered an orphan nuclear receptor (53). Cloning and characterization of the mosquito lipophorin gene will be an important step toward further understanding of the possible role of HNF-4 in regulating lipogenesis at the gene level.
In contrast to AaHNF-4a, the AaHNF-4b mRNA accumulates during the previtellogenic period and then drops to a lower level that is maintained through the vitellogenic synthetic phase. The identification of HNF-4 binding sites in the putative regulatory regions of two yolk protein precursor genes, A. aegypti vg (26) and vcp (27), suggests the possible role of HNF-4b in regulating the expression of these genes. HNF-4 binds as a homodimer to direct repeats of the sequence AG-GTCA, with a one-nucleotide spacer between the two half-sites (10). The putative HNF-4 motif in the A. aegypti vg gene closely corresponds to the consensus HNF-4 binding site (10). The putative motif in the A. aegypti vcp gene is more distinctive in that it has two nucleotides that separate the two half-sites, previously reported only for the erythropoietin gene (54). Further studies utilizing cell transfection assays and insect gene transformation should elucidate the role of these HNF-4 binding sites in regulating the expression of these mosquito yolk protein genes.
Synthesis of the yolk protein precursors reaches its maximal levels between 24 -27 h PBM and then decreases dramatically (15). It is at this time that the level of the AaHNF-4c transcript increases relative to the two other AaHNF-4 transcripts. This pattern of expression of the nuclear factor isoform lacking the DNA-binding domain raises a possibility of its role as a transcriptional repressor involved in switching gene regulation in the mosquito fat body. A number of proteins with abolished DNA-binding function have been identified, including E75B (42), SHP (55), Id (56), and ERbA␣ p30 and p27 (57); all belong to families of dimeric transcription factors. We propose that the AaHNF-4c isoform can act as a dominant negative factor by forming heterodimers with other AaHNF-4 isoforms, which are incapable of DNA binding and subsequent transcriptional activation. In support of this hypothesis, we have shown that in vitro AaHNF-4c can abolish the binding of AaHNF-4a, AaHNF-4b, and even rat HNF-4 to target DNA sequences. The ability of the mosquito HNF-4c isoform to inhibit the formation of binding complexes between the rat HNF-4 and the HNF-4 response element clearly demonstrates dimerization between these heterologous HNF-4 factors.
Furthermore, transcription studies in mammalian cells support the idea that the mosquito HNF-4c isoform can act as a dominant negative via dimerization with wild-type HNF-4. The precise role of the AaHNF-4c isoform in regulation of the fat body-specific gene expression in the mosquito remains to be elucidated. However, the discovery of an HNF-4 isoform that lacks the DNA binding domain and serves as a transcriptional repressor has expanded our view on potential functions of this important transcription factor.